When modelling classes, what is the preferred way of initializing:
And what would be the considerations for using either of them?
In certain situations, I prefer having a factory method which returns null if the object cannot be constructed. This makes the code neat. I can simply check if the returned value is not null before taking alternative action, in contrast with throwing an exception from the constructor. (I personally don't like exceptions)
Say, I have a constructor on a class which expects an id value. The constructor uses this value to populate the class from the database. In the case where a record with the specified id does not exist, the constructor throws a RecordNotFoundException. In this case I will have to enclose the construction of all such classes within a try..catch block.
In contrast to this I can have a static factory method on those classes which will return null if the record is not found.
Which approach is better in this case, constructor or factory method?
From page 108 of Design Patterns: Elements of Reusable Object-Oriented Software by Gamma, Helm, Johnson, and Vlissides.
Use the Factory Method pattern when